Managerial Compensation and Stock Price Informativeness

Transcription

1 Managerial Compensation and Stock Price Informativeness Benjamin Bennett, Gerald Garvey, Todd Milbourn, and Zexi Wang Draft: January 28, 2017 (Preliminary; please do not quote) Abstract This paper studies the effect of stock price informativeness (SPI) on the level, complexity, and efficiency of executive incentive pay. A more informative stock price provides principals more information on agents effort. This implies a lower amount of total and equity-based pay because less risk needs to be imposed on executives to induce a given level of effective incentives. This argument assumes that executive compensation is chosen efficiently; we find our effects are stronger in firms with better governance. We also find that the stock price informativeness reduces pay complexity; firms with more informative stocks have shorter pay duration, higher percentage of cash pay, and are less likely to use compensation consultants. In a multi-task setting, even a highly efficient and informative stock price can inefficiently aggregate information across tasks; consistent with this we find weaker results for multidivisional firms. Lastly, SPI increases the efficiency of compensation through its effect on the pay performance sensitivity, CEO perquisites, and pay for luck/skill. Fisher School of Business, Ohio State University, 840 Fisher Hall, 2100 Neil Ave, Columbus OH 43210, USA University of New South Wales, Australia Business School, College Rd, Kensington NSW 2052, Australia John M. Olin School of Business, Washington University, Campus Box 1133, 1 Brookings Dr, St. Louis, MO 63130, USA University of Bern, Institute for Financial Management, Engehaldenstrasse 4, CH-3012 Bern, Switzerland 1

2 Introduction In recent years, the level of executive compensation has soared, and compensation packages have become more and more complex. In 2015, the median CEO total compensation of S&P 500 firms reached 10.4 million US dollars, compared to 7 million US dollars in Additionally, the complexity of compensation packages continues to increase as well. For example, many pay plans include contingencies or long-term performance goals that make them unwieldy, as commented by Bengt Holmstrom in his Nobel Prize press conference and echoed by the concerns of investors in practice 1. Executive compensation is designed to address the agency problem (Jensen and Meckling 1976) and the information asymmetry between agents and principals plays a fundamental role (Holmstrom 1979; Shavell 1979). Different information sources decrease information asymmetry, we focus on stock prices which incorporate the latest firm specific information, including that reflecting skills and efforts of managers. Stock price informativeness improves the observability of managerial efforts, which suggests that the informativeness of stock prices affect executive compensation design. In this paper, we study the effect of stock price informativeness on the executive compensation. Stock price informativeness (SPI) can affect the level of the executive compensation. Stock prices serve as a signal of managerial skill and effort, and the informativeness measures the quality of the signal. New information on firms is incorporated into stock prices through informed trading, and all economic units can learn from the stock prices (Hayek 1945; Grossman and Stiglitz 1980; Kyle 1985). Existing literature shows that the informativeness of stock prices has real effects on corporate decisions, such as investment, and corporate 1 See the article in Financial Times at 2

3 governance (Bond, Edmans, and Goldstein 2012; Chen, Goldstein, and Jiang 2007; Ferreira, Ferreira, and Raposo 2011). Regarding the compensation, more informative stock prices can decrease the information gap between agents and principals, and increase the observability of agents effort. For example, higher talent and more effort of agents increase the probability of value-maximizing decisions and actions, which are more likely to be reflected by higher stock prices when the prices are more informative. The higher obervability of agents efforts decreases risk sharing towards the first best solution (Holmstrom 1979). It implies that less equity-based pay is used in the compensation for risk sharing. Accordingly, the risk-averse agents require less total compensation, which benefits the principals. Corporate governance plays an important role in the effect of SPI on the level of total compensation. Lower compensation benefits the principals, but it hurts the managers income. It is intuitive that the negative effect of SPI on executive compensation should be stronger for firms with better governance. For firms with weaker governance, it is more difficult to cut the compensation even if the observability is improved by the SPI. We find the evidence for the complementary effect of the corporate governance. SPI has significant effect on compensation only in firms with better governance. We further investigate the variations of the SPI effect on equity-based compensation along firm characteristics. We consider the roles of financial constraints, product market competition, and corporate diversification. Financially constrained firms suffer from inaccessible external financing, which brings more uncertainty on firms growth and makes the firm s equity less attractive to managers. As a result, managers of financially constrained firms are more reluctant to accept equity as part of their compensation. If they have to accept equity, 3

4 they will charge a premium, which makes the pay more costly for the principals. This gives principals a stronger motive to decrease equity-based pay for a given level of SPI. We find that the negative effect of SPI on the equity-based pay is stronger for financially constrained firms. The product market competition impacts SPI s effect on the equity based pay. Product market competition can also make the managerial effort more observable to principals (Hart 1983). It implies that SPI and competition are substitutes. We expect the SPI effect on equitybased pay to be stronger for firms facing less product market competition. We measure competition by the product market fluidity (Hoberg, Phillips and Prabhala 2014), product similarity, and HHI (Hoberg and Phillips 2016). The empirical results of all three measures consistently verify that low competition in the product market amplifies the SPI effect on equity-based pay. Corporate diversification can also affect the use of equity-based pay. Equity based pay is less efficient for some firms than for others. For example, Paul (1992) shows that equity based pay can be inefficient when firms have multiple projects. The reason is that the equity based pay puts higher weights on the projects where the managers effort is less observable. The stock price reflects the value of the firm, not the added value by the managers. It implies that the equity based pay is more likely to be inefficient when firms have more projects. We expect the negative effect of SPI on equity pay to be stronger in the firms with more business segments (more diversified). Specifically, the interaction between SPI and the number of segments is expected to have negative impact on the percent of equity based pay in the total compensation. 4

5 Stock price informativeness (SPI) can decrease the complexity of compensation. Recently, complex compensation packages make it difficult for investors to understand and disentangle the real incentives provided to the managers, especially by the long-term incentive plans. SPI improves the transparency of the managers effort, which makes the complex compensation less necessary. Suppose the efforts of managers are totally transparent to the principals, the managers can be paid wholly in cash for their effort, which is the simplest compensation. Compensation complexity is measured by the following four dimensions: the percent of cash pay in the compensation, the duration of the compensation (Gopalan, Milbourn, Song, and Thakor 2014), the dispersion of components in the total compensation (Albuquerque, Carter, and Lynch 2015), and the usage of compensation consultants. It is intuitive that a higher percent of cash pay implies less complex compensation. Pay duration reflects the vesting periods of different pay components. Shorter duration implies lower complexity. Lower dispersion indicates that only a few components of the compensation dominate the pay package, and the compensation package is less complex. Sometimes compensation consultants are blamed to be responsible to the complex compensation 2. On the one hand, the consultants may have an incentive to design complex compensation to justify their existence. On the other hand, firms with more complex compensation demands are more likely to hire compensation consultants. The use of compensation consultants should be positively correlated with the compensation complexity. The results of all four measures show that SPI decreases the complexity of executive compensation

6 We further focus on the performance-based pay, which increases the complexity of compensation. Performance-based pay makes the manager s compensation conditional on certain performance metrics, such as an earnings or sales target. It has been more and more popular in US firms (Bennett, Bettis, Gopalan, and Milbourn 2016; Li and Wang 2016). The main purpose of the performance-based pay is to provide strong incentive to managers. However, the incentive instruments are costly. For example, managers can require risk premium and liquidity premium for equity-based pay or longer pay duration. SPI improves the observability of managers effort, and thus less incentive instruments are needed to maintain a certain level of managerial effort. We expect SPI decreases the demand for performance-based pay, especially the pay based on the long-term accounting performance. We find evidence that SPI decreases the proportion of performance based pay in the total compensation, especially the pay based on the long-term performance, and the pay based on the accounting performance. 3 Besides the level and complexity of the compensation, we also investigate the effect SPI has on compensation efficiency. As more information is available to firms with higher SPI, we expect SPI to increase the efficiency of the compensation. Our results in previous sections show that SPI decreases the level of equity-based pay, which is a typical incentive instrument. After investigating how SPI affects the level of the incentives used in compensation, we seek to understand how SPI affects the incentives (as opposed to their levels). Will the incentive relationship between equity and managerial effort also be negatively affected by SPI? The answer is no. We find that SPI actually increases the pay performance sensitivity (delta). As more informative stocks serve as a higher quality signal, compensation can rely on stock prices 3 One other cost of performance-based pay tied to specific accounting goals is that it is related to manipulation of firm characteristics (see Bennett, Bettis, Gopalan and Milbourn, 2017). 6

7 in a more aggressive way. For example, firm can use more option compared to stock pay. Given the same pay amount, option pay provides stronger incentive than the stock pay does. We find the ratio of option pay to equity pay is positively affected by the SPI. It may explain why SPI leads to lower level of equity pay and stronger incentive. Another channel by which the SPI affects incentives is through performance-based pay tied to stock prices. Higher SPI meansuseful information is more likely to be contained in the stock price. The informativeness principle implies that stock prices should be used more aggressively in the compensation design for firms with higher SPI (Holmstrom 1979). We expect SPI to increase the reliance on stock prices in the performance-based pay, where the managers get the certain amount of pay only if the firm s stock return is higher than a specified level. We find that SPI does increase percentage of stock price-related performance-based pay. In this way, SPI increases the sensitivity of pay to the stock performance, and has positive effect on the efficiency of the compensation. Stock price informativeness can improve the pay efficiency by decreasing the managerial perquisites. Executive perquisites are generally thought of as an agency problem to misuse firm resources (Grossman and Hart 1980, Jensen and Meckling 1976, Jensen 1986). Inappropriate perquisites can have significant effects on stock prices when they are known by outside investors (Yermack 2006). Higher SPI implies such perquisites are more likely to be known by outsider investors, and makes CEOs less likely to enjoy costly perquisites. We find that SPI has significantly negative effect on the CEO perquisites, which improves the efficiency of compensation. 7

8 SPI can also improve efficiency through its effects on pay for luck/skill. Efficient compensation should reward managers for their effort and skills, not for the outcomes outside their control (luck). We measure luck by the firm s return predicted by its industry return, and measure skill by the residual of the same following Garvey and Milbourn (2006). We find that pay for luck concentrates in the firms with low SPI and pay for skill concentrates in the firms with high SPI. These results indicate that firms with high SPI have more efficient executive compensations. Our contributions are as follows. First, the paper contributes to the literature on executive compensation. It finds that the stock price informativeness affects the level, complexity, and efficiency of compensation. Second, it contributes to the literature on the real effect of financial markets. It shows a channel through which the stock market can affect managerial compensation, which in turn has comprehensive effects on corporate behavior. Third, this paper contributes to the literature on informed trading. Stock price infomativeness is positively affected by the informed trading, and decreases the compensation. It implies the informed trading may have positive effect on principals interests. The remaining sections are organized as follows. Section 2 introduces the measures of stock price informativeness. Section 3 describes the data sources and the sample. Section 4 shows the effect of SPI on the level of compensation and the complementary effect of corporate governance. Section 5 shows how the SPI effect on equity based pay varies along firm characteristics. Section 6 shows the effect of SPI on the complexity of compensation. Section 7 shows the SPI effect on the performance based pay. Section 8 shows how SPI affects the efficiency of compensation. Section 9 concludes. 8

9 2. Measures of stock price informativeness 2.1 Stock price informativeness and corporate decisions Prices can aggregate the information owned by different units in the economy (Hayek 1945). In fact, production, aggregation and communication of information are some of the most important roles of modern financial markets. When different economic participants own private information related to firms, they can make profits through speculative trading of corporate stocks. When doing so, their private information is incorporated into the stock prices (Grossman and Stiglitz 1980; Kyle 1985). Then all participants can share the private information through stock prices, and the information asymmetry among different parties is decreased. This information sharing plays an important role in the real effects financial markets have on the economy (Bond, Edmans, and Goldstein 2012). In particular, the amount of private information in stock prices can affect corporate decisions. In the literature, the amount of the private information in the stock prices is denoted as the stock price informativeness (SPI). Existing evidence shows that SPI can affect corporate activities such as investment, cash savings, and corporate governance. Chen, Goldstein, and Jiang (2007) find that SPI has strong effect on corporate investment. The investment of firms with higher SPI is more sensitive to stock prices, because the managers can learn more from stock prices with higher SPI. Consistently, Fresard (2012) finds that SPI affects corporate cash savings. There is also evidence that SPI affects corporate governance. Ferreira, Ferreira, and Raposo (2011) find that SPI decreases the corporate board independence. SPI makes managers actions more observable when private information is 9

10 released through the stock market, and directors could use such information as an input to their monitoring task. They find SPI and board independence are substitutes. 2.2 Measures of stock price informativeness Following the literature, we use two measures of stock price informativeness, which are annual level measures based on high frequency tick size trading, or daily trading activities. The first is the probability of information-based trading (PIN), which is constructed based on a market microstructure model (Easley, Hvidkjaer, and O Hara 2002). The logic is that when there is more informed trading in one stock, new information is more likely to be incorporated into that stock s price, which improves the stock price informativeness. High PIN means high stock price informativeness. The second is the stock price nonsynchronicity (PSI), which is the firmspecific return variation (Durnev, Morck, and Yeung 2004). The logic is that when there is more firm-specific information in the stock price, the stock return is less correlated with market and industry returns. High PSI means high stock price informativeness. Both measures are widely used as stock price informativeness measures in the relevant literature. We use two measures of stock price informativeness. The first is the probability of information-based trading (PIN), which is constructed based on a market microstructure model (Easley, Hvidkjaer, and O Hara 2002). The logic is that when there is more informed trading in one stock, new information is more likely to be incorporated into that stock s price, which improve the stock price informativeness. High PIN means high stock price informativeness. The second is the stock price nonsynchronicity (PSI), which is the firm-specific return variation (Durnev, Morck, and Yeung 2004). The logic is that when there is more firm-specific information in the stock price, the stock return is less correlated with market and industry returns. High PSI 10

11 means high stock price informativeness. Both measures are widely used as stock price informativeness measures in the relevant literature Probability of information-based trading (PIN) PIN is the measure of the probability of information-based trading. Suppose that on a day the new information appears with probability α, and with probability δ the news is bad, and with probability of 1 δ the news is good. Then the probability of no news on a day is 1 α. The trading orders follow Poisson distributions. Uninformed traders (with on information advantage) trade no matter new information arrives or not. The arrival rate of uninformed buy (sell) order is ε b (ε s ). The traders with private information only trade when there is new information appears, and the arrival rate is μ. The informed trader will only buy if the news is good and only sell if the news is bad. Given these parameters (α, δ, μ, ε b, ε s ), the probability of information-based trading is PIN = α μ α μ + (ε b + ε s ), where the denominator is the arrival rate for all orders and the numerator is the arrival rate of informed orders. The parameters are estimated by method of maximum likelihood. On day i, we observe the number of buy orders B i and the number of sell orders S i. Denote the Poisson distribution function as P(k; λ) = e and λ is the arrival rate. Then the likelihood function for a trading day is L(α, δ, μ, ε b, ε s B i, S i ) λ λk, where k is the number of arrivals k! = (1 α) P(B i ; ε b ) P(S i ; ε s ) + α δ P(B i ; ε b ) P(S i ; μ + ε s ) + α (1 δ) P(B i ; μ + ε b ) P(S i ; ε s ) 11

12 Assume the tradings across different days are independent. Then the likelihood function within a year is I V = L(α, δ, μ, ε b, ε s B i, S i ), i=1 where I is the number of trading days in a year. Based on TAQ data and Lee and Ready (1991) algorithm, we calculate the number of daily buy and sell orders of a stock. Then use the maximum likelihood method to calculate the parameters (α, δ, μ, ε b, ε s ) based on the data in a year. In turn, PIN can be calculated for a stock in a given year Stock price nonsynchronicity (PSI) The stock price nonsynchronicity, PSI, is a measure of stock price informativeness based on R 2, in line with Roll (1988). In the following regression, the stock return is decomposed into the systematic part explained by market return and industry return, and a firm-specific residual variation. When there is relatively more firm-specific variation, the return co-moves less with market return and industry return, and in turn smaller R 2 (or equivalently larger 1 R 2 ). r j,i,t = β j,0 + β j,m r m,t + β j,i r i,t + ε i,j,t, where j is for firm j, i is for industry i, and t is for day t, r j,i,t is the stock return of firm j in industry i (three-digit SIC) on day t, r m,t is the value weighted market return on day t, and r i,t is the value weighted industry return on day t. The weight is market capitalization. When calculating the market and industry value weighted returns for firm j, the return of firm j is excluded to prevent spurious correlations between firm and industry returns in industries that contain few firms. 12

13 The regression is run for each firm j within a year, and the R 2 of the regression is used to construct PSI j for stock j in a given year as follows. PSI j = ln 1 R j 2 Therefore, PSI j is a logistic transform of 1 R 2 j, which is to address the skewness and boundedness of 1 R 2 j. Stock price is more informative when a stock becomes less correlated 2 with the market and industry returns, i.e. smaller R j and larger PSI j. 3. Data and Sample R j 2 Executive compensation data are from Execucomp and IncentiveLab. 4 Corporate accounting data are from Compustat. We use TAQ data to calculate PIN and daily stock file of CRSP to calculate PSI. Institutional ownership and blockholder data are from Thomson Reuters 13F. E-index is from RiskMetrics. Competition variables used are from Hoberg-Phillips data library. 5 Our sample only includes firms with non-missing compensation data and non-missing stock price informativeness (we require at least one of PIN or PSI for a firm-year to be included in our sample). PIN is first available in 1993 as that is the first year TAQ data is available. In our analysis, we use the average PIN and PSI of the previous three years (at least one non-missing value in the previous three years). We use such a backward looking approach to help alleviate reverse causality concerns. Our sample is from 1994 to 2015 and includes 35,005 firm-year observations. 4 For a more thorough discussion of the IncentiveLab data see Bettis, Bizjak, Coles and Kalpathy (2016) or Bennett, Bettis, Gopalan and Milbourn (2016). 5 The competition data and a descriptions of same can be found here: 13

14 The control variables used in the main tests are the log of book assets, R&D/assets, return on assets (ROA), Tobin s Q, cash/assets, debt/assets and sales growth. The definitions for all variables can be found in the Appendix. The summary statistics for the main variables are reported in Table 1. The means of PIN and PSI are 0.18 and 1.28, which are in line with recent, previous studies. 6 The means for CEO total, cash, and equity compensation are also similar to those in recent studies at $2.7M, $0.9M, and $1.0M. 7 Empirical analysis We now discuss and present the results of our main empirical tests. Stock price informativeness is related to the level, complexity, and efficiency of executive compensation. Our empirical analyses verify the relevant hypotheses. 4. Level of compensation 4.1 Level of compensation and SPI We start with the tests for the level of compensation. Stock price informativeness (SPI) decreases the information asymmetry between principals and agents, and increases the observability of agents effort. Such transparency is supposed to decrease the necessity of risksharing such as equity-based pay (Holstrom 1979), and in turn decrease the compensation to the risk-averse managers. The SPI is expected to have negative effects on both total compensation and equity-based compensation. The total compensation (equity based 6 See Chen, Goldstein, and Jiang (2007) and Ferreira, Ferreira, and Raposo (2011). 7 There are many compensation studies, but a prominent, recent overview is Murphy (2012). 14

15 compensation) is measured by the logarithm of total compensation (equity-based compensation) of CEOs. We use 2 measures of SPI in all of our tests: the probability of insider trading (PIN) and the stock price non-synchronicity (PSI). 8 The measures are first calculated for each firm in a year. Then we use the average PIN or PSI across the previous 3 years as the SPI of a firm in a given year. This moving average with 3-year window has at least three advantages. First, it avoids the effect of the large change of SPI in any specific year, and better measures the average level of SPI. Second, using SPI in previous years alleviates the endogeneity concern. Third, executive compensation contracts generally vest over multiple years with three years being the mode (see Bettis et al, 2016), thus using SPI over multiple previous years can be considered when signing a new contract. The results are reported in the Table 2. All regressions include industry and year fixed effects, and robust standard errors are clustered at the firm level. Columns 1 and 2 of Table 2 report the results on equity based pay. The equity based pay includes the stocks and options in the compensation package. The results show that the coefficients of PIN and PSI on equity pay are both negative and statistically significant at 1% level. The economic effects are also significant. One standard deviation increase of PSI (PIN) decreases equity pay by 27% (8%). Compared to the effects on total compensation, the effects on equity-based pay are even stronger. Columns 3 and 4 report the results on total pay. Both PIN and PSI have negative effect on total compensation, and the effects are statistically significant at 1% level. The one standard 8 These measures and their construction are discussed in detail in Section 2. 15

16 deviation increase of PSI (PIN) decreases total compensation by 7% (3%). The results verify the negative effect of stock price informativeness on executive compensation. In the Columns 5 and 6, we further investigate the effect of SPI on the cash pay, which includes the salary and bonus. The coefficients of PIN and PSI are both negative, but only the coefficient of PIN is statistically significant. It indicates that SPI also has negative effect on the cash compensation. But the effect is not as strong as the effect on equity based pay. It is consistent with that cash pay, such as salary, may be more rigid than the equity pay, which makes the equity pay the more variable part of compensation. 4.2 Corporate governance and complementary effect In this section, we study the SPI effects on compensation of firms with different levels of corporate governance. High SPI decreases the information asymmetry, which theoretically should have a negative effect on optimal compensation. Lower total compensation benefits principals. However, it hurts the managerial pocketbook. The negative effect of SPI on total compensation is expected to be stronger in firms with better governance, where managers have less power with regards to rent extraction. To test this hypothesis, we divide our sample into strong and weak governance groups. We use two measures of corporate governance: one is the E-index (Bebchuk, Cohen, and Ferrell 2009), and the other is the presence of blockholder. The blockholder is an institutional investor holding more than 5% of shares outstanding. Large shareholders have stronger monitoring motive and improve the corporate governance (Shleifer and Vishny 1986). We define the strong-governance group as the firms with at least one blockholder, and weak-governance group as firms without blockholder. Alternatively, based on E-index, the strong (weak)- 16

17 governance group is defined as the firms with E-index below (above) the median in a year. We then regress the total compensation on the SPI measures separately for strong governance and weak governance groups. The regression settings and control variables are the same as those in Table 2. The results are shown in Table 3. The results based on the presence of E-index (blockholder) are in the top (bottom) panel of Table 3. We find that SPI mainly effects firms with strong governance. For example, in the top panel of Table 3, PIN is significant in the strong-governance group (E-index below the median), but not the weak governance group. It indicates the complementary role of corporate governance: together with strong governance, SPI effects total compensation and benefits principals. 5. Equity based pay and SPI In this section, we investigate how the SPI effect on equity-based pay varies with firm characteristics. We consider three firm characteristics: financial constraints, product market competition, and diversification. The cross-sectional variations of these characteristics are expected to influence the SPI effect on equity-based pay. 5.1 Financial constraints Financial constraints affect the relationship between SPI and equity-based pay. The previous evidence shows that firms with higher SPI use less equity-based pay. The cost of equity-based pay can be higher in some firms than in others. Given the level of SPI, we expect firms where the equity-based pay is more costly to have stronger motives to cut equity-based pay because it provides a relatively larger benefit to principals. Financial constraints increase the cost of equity-based pay. The reason is as follows. Financial constraints mean that it is more 17

18 difficult for a firm to access external financing when a valuable investment opportunity appears. To a financially constrained firm, ceteris paribus, its prospects are less likely to be realized, which makes managers more reluctant to accept equity-based pay and/or require a higher premium for accepting equity-based pay. Therefore, the cost of equity-based pay is higher for financially constrained firms. In our main results we find that higher SPI leads to less risk-sharing and higher incentives for the principal to decrease equity-based compensation. Consistent with those results, here we find that the principal will decrease the equity-based compensation to a larger degree when it is more expensive as it is for financially constrained firms. We expect that financial constraint amplifies the negative effect of SPI on equity based pay. 9 We use four measures of financial constraints: firm size, bond rating, commercial paper rating, and Whited-Wu index (Whited and Wu 2006). Firm size is measured by the logarithm of total assets. Small firms are more likely to be financially constrained. Firms having no access to the public debt market are more likely to be financially constrained. We define a dummy variable fc_bond (fc_cp), which equals to 1 if the firm has no S&P bond (commercial paper) ratings, and 0 otherwise (Almeida, Campello, and Weisbach, 2004; Denis and Sibilkov, 2010). The Whited-Wu index is constructed as in Whited and Wu (2006), and firms with larger Whited- Wu index are more likely to be financially constrained. Following the literature on financial constraints, we drop the financial and utility firms (Denis and Sibilkov, 2010) from these tests. 9 In this section we focus on how financial constraints amplify the effect of SPI on equity-based pay, rather than the effect of financial constraints on equity-based pay. Financially constrained firms might rely more on equity-based pay because of low liquidity. However, the literature shows mixed evidence on it. For example, Core and Guay (2001) show that financially constrained firms use more stock options to pay nonexecutive employees. Ittner, Lambert, Larcker (2003) shows that firms with higher cash flows use more equity-based pay. 18

19 To study the amplification effect of financial constraints, we focus on the interaction between the SPI and the financial constraint measures. The results are shown in Table 4. We have four financial constraint measures and two SPI measures, which provide eight interaction items in total. Accordingly, in all eight regressions, the coefficients of the interaction items indicate that financial constraint amplifies the effect of SPI on equity based, and six out of the eight coefficients are statistically significant at least at 10% level. Columns 1 and 2 show the results for firm size as the financial constraint measure. The interactions have positive coefficients, which imply that the negative effect of SPI on equitybased pay is weaker for larger firms (financially unconstrained). Put differently, the effect is stronger for smaller firms, which are more likely to be financially constrained. The coefficients on the interaction terms in Columns 3 to 8 are all negative, which shows that the financial constraint amplifies the effect of SPI on equity based pay. For example, Columns 5 and 6 shows the results for the financial constraint dummy based on commercial paper ratings. Its interactions with PIN and PSI are both negative and statistically significant at 1% level. The results in Table 4 provide strong support to the hypothesis that financial constraints amplify the negative effect SPI has on equity-based pay. 5.2 Product market competition The effect SPI has on equity-based pay is expected to be stronger for firms with less product market competition. Product market competition makes the manager s efforts more observable to principals because competition makes firm performance more transparent (Hart 1983). Managerial effort in firms with less competition is less transparent to principals. This implies SPI plays a more important role in firms with less competition. Said another way, SPI is a 19

20 substitute for competition. Both of, SPI and competition, make the manager s efforts more visible to the principal. We expect that the effect SPI has on equity-based pay should be stronger in firms with weaker product market competition (substitution effect). In firms with low competition, the principal cannot learn about the agent s effort through the competition channel, thus being able to do so through the SPI channel becomes more important. We use three different measures of product market competition. The first is the product market fluidity (Hoberg, Phillips and Prabhala 2014), which measures how intensively the product market around a firm is changing in each year. The second is industry HHI, and the third is the industry product similarity (Hoberg and Phillips 2016), where the industry is the 10-K text-based network industry classification (TNIC) 10. We define a dummy variable Low_Fluidity (Low_similarity), which equals 1 if the fluidity (product similarity) is below the median in a year, and 0 otherwise. We define a dummy variable High_HHI, which equals 1 if the HHI is above the median in a year, and 0 otherwise. The value 1 for each of these three dummies indicates that a firm has relatively low product market competition. To investigate how product market competition influences the SPI effect on equity based pay, we focus on the interactions between the competition dummies and the SPI measures. We expect the coefficients of the interaction terms to be negative, which indicates that the negative effect of SPI on equity-based pay is amplified in firms with less product market competition. The results are shown in Table 5. The regression results provide strong support to our hypothesis on competition. The coefficients of all six interaction items are 10 We are grateful to Hoberg and Phillips for providing the measures at the Hoberg-Phillips data library. 20

21 significantly negative. For example, Column 1 shows the result of the interaction between Low Fluidity and PIN. The coefficient of the interaction is and statistically significant at 1% level. It indicates that the effect of PIN on equity-based pay mainly concentrates in low competition firms, while the effect in high competition firms is trivial. The results for PSI also confirm that the SPI effect on equity-based pay is stronger for firms with less product market competition. 5.3 Diversification Corporate diversification can influence the effect SPI has on equity-based pay. In a multi-project firm, Paul (1992) shows that equity-based pay makes compensation less efficient because equity-based pay puts higher weights on projects where the managers effort is less observable. The reason is that the stock price reflects the value of the firm, not the value added by the managers. This implies that the equity-based pay is more likely to be inefficient when firms have more projects. Therefore, we expect the negative effect of SPI on equity-pay to be stronger in the firms with more business segments (more diversified). Specifically, the interaction between SPI and the number of segments is expected to have a negative impact on the percent of equity-based pay in the total compensation. The results are shown in Table 6. Columns 1 and 2 show that both PIN and PSI reduce the percent of equity-based pay. The result for our tests which include the number of segments as a measure of diversification are shown in Column 3 (4). The coefficient of the interaction between PIN (PSI) and the number of segments is negative and statistically significant at 1% (5%) level. It indicates that more diversified firms use less relatively equity based pay in the compensation. Diversification amplifies the effect of SPI on the relative equity based pay in the compensation. 21

22 6. Complexity of compensation Stock price informativeness (SPI) can decrease the complexity of the compensation. In recent years, the compensation package becomes more and more complex. This additional and increasing complexity makes it more difficult for investors to understand managerial compensation structure and design. However, as SPI makes the managers effort more observable, the demand for complex compensation packages is lower. In an extreme case, if a manager s efforts are perfectly observable, he should be paid only using cash as his salary is based only on working hours with no complex incentive structure required. As such, we expect SPI to decrease compensation complexity. We use four proxies of compensation complexity. The first is the percentage of cash pay in the compensation. Compared to equity-based pay, cash pay is a simpler and more direct method to pay managers. Intuitively, a higher percentage of cash pay is relatively less complex. The second proxy is the duration of executive compensation, which reflects the vesting periods of different pay components (Gopalan, Milbourn, Song, and Thakor 2014). Generally, the longer the pay duration is, the more complex the compensation package tends to be. The third proxy is TDC1 complexity, which is the dispersion of total compensation among different components of pay (Albuquerque, Carter, and Lynch 2015). First calculate the Herfindahl Index of the proportions of each component of total compensation. A high Herfindahl Index indicates that only a few components dominate the pay package. Then TDC1 complexity, the dispersion, is defined as 1 minus this Herfindahl Index. A high dispersion means that more components are used and plan relative important role, which indicates high complexity of the pay. The fourth proxy is the usage of compensation consultants. Compensation consultants are often criticized 22

23 to be responsible for the level complexity in compensation packages. We define a dummy variable, which equals to 1 if a firm uses a compensation consultant and 0 otherwise. Firms using compensation consultants are more likely to have more complex compensation. The results are reported in Table 7. Columns 1 and 2 show the relationship between SPI and the percentage of cash pay. Both PIN and PSI have positive and statistically significant coefficients. They indicate that SPI increases the use of cash as the payment method in the compensation, which makes the compensation less complex. Columns 3 and 4 display results for the relationship between SPI and pay duration. Column 3 shows that PIN has a negative effect on the pay duration, and the coefficient is statistically significant at 1% level. Column 4 shows a consistent result for PSI. These results indicate that SPI decreases pay duration, which in turn makes compensation less complex. Furthermore, executives with longer pay durations have higher total compensation (Gopalan, Milbourn, Song, and Thakor 2014), the evidence here is consistent with our previous results that SPI decreases the executive compensations. Columns 5 and 6 show the results for the dispersion of total compensation (TDC1 complexity). Both PIN and PSI decrease the dispersion, and the coefficients are statistically significant at 1% level. They provide strong support that SPI decreases the compensation complexity. Columns 7 and 8 present the results of logistic regressions which test for a relationship between SPI and the usage of compensation consultants. Both PIN and PSI have negative coefficients, which indicate that firms with high SPI are less likely to use compensation consultants. This is consistent with SPI decreasing compensation complexity. 7. Performance based pay 23

24 US firms increasingly tie managerial pay to explicit performance metrics, such as a specific earnings or sales target (Bennett, Bettis, Gopalan, and Milbourn 2016; Li and Wang 2016). The performance-based pay may offer strong incentives, but it can also have a dark side. In general, investors complain that performance based pay, especially the long-term performance-based pay, makes the compensation too complex to understand. 11 Performancebased pay can also incentivize managerial manipulation of firm characteristics to achieve these performance targets, which is not consistent with maximizing shareholder value (Bennett, Bettis, Gopalan, and Milbourn 2016). Because SPI improves the observability of managerial effort, fewer incentive instruments are needed in the optimal compensation package. Thus, we expect SPI to reduce the demand for performance-based pay, especially long-term or accounting performance-based pay. To investigate the effect of SPI on the performance-based pay, we consider the proportion of total compensation that is performance-based, the proportion of long-term performance-based pay, the proportion of the pay based on accounting performance, and finally, the proportion of the pay based on long-term accounting performance. The long-term here refers to the pay that vests in more than one year. We define a dummy High_PIN (High_PSI), which equals 1 if the PIN (PSI) is in the top quartile and 0 if the PIN (PSI) is in the bottom quartile. The High_PIN (High_PSI) is expected to have negative effect on the performance based pay. The results are shown in Table 8. Columns 1 and 2 of Table 8 show the results for the proportion of performance based pay in the total compensation. Column 1 (2) shows that the coefficient of High_PIN (High_PSI) is

25 negative and statistically significant at 10% (5%) level. It indicates that firms with higher PIN have smaller proportion of performance based pay. Columns 3 and 4 show the results for the proportion of long-term performance-based pay. Both High_PIN and High_PSI are negative and significant at 5% level which indicates that the long-term performance-based pay is significantly reduced by SPI. Columns 5 and 6 focus on the proportion of accounting performance-based pay. The coefficients of both SPI measures are negative and statistically significant at 5% level, which indicates that when SPI is higher, the reliance on accounting performance in compensation is lower. Columns 7 and 8 shows the results for the proportion of long-term accounting performance-based pay. The coefficients of the SPI measures are once again negative and statistically significant at 5% (10%) level. In the unreported analysis, we consider the proportion of short-term accounting performance based pay. It turns out that the coefficients of High_PIN and High_PSI are both insignificant at conventional levels. It implies that the effect of SPI on accounting performance based pay concentrates in the long-term accounting performance based pay. All evidence in Table 8 indicates that SPI decreases the use of performance based pay, especially the long-term and accounting performance-based pay. These findings are consistent with the results in the previous section that SPI decreases the complexity of compensation. 8. Efficiency of compensation SPI decreases total and equity-based compensation, which may benefit the principals. Does lower equity-based pay have negative effects the incentives of managers? In this section, we study the effect SPI has on compensation efficiency. First, we investigate how delta of the compensation is affected, and the potential reasons. Second, we investigate how the perquisite 25

26 compensation is affected by SPI. Third, we study how SPI affects pay for luck/skill of the compensation. 8.1 Incentive design Compensation is designed to align the interest of agents and principals. Equity-based pay generally serves as an incentive instrument. Equity-based pay incentives can be affected by the level and the structure of equity-based pay. The level of the equity-based pay generally strengthens the incentive. Similarly, the structure of the equity-based pay also matters because for the same dollar amount, different components can provide different or stronger incentives. For example, stock options can provide stronger incentives than the same dollar amount of stock. We know, from earlier tests, that SPI decreases the level of equity-based pay. But we are, as of yet, unsure of its effects on the equity-based pay structure. We first investigate the effect of SPI on delta, or pay-performance sensitivity. The results are shown in Columns 1 and 2 in Table 9. Column 1 shows that the coefficient of PIN is positive and statistically significant at 1% level. And the result on PSI in column 2 is consistent. These results indicate that SPI increases the pay-performance sensitivity, which is consistent with a positive impact on equity incentives. Even though the level of equity-based pay is decreasing with SPI, the delta is increasing with SPI. We expect SPI to increase the usage of more aggressive equity incentive instruments. For example, more options relative to the stocks may be used. If this were the case, then SPI would simultaneously increase pay-performance sensitivity and decrease the level of equitybased pay. These two simultaneous SPI effects would result in more efficient compensation. We use the ratio of option pay to stock pay as a measure of the aggressiveness of the equity- 26

27 based pay structure. The results of these tests are shown in Columns 3 and 4 in Table 9. The results show that both PIN and PSI have significantly positive effects on the ratio of option pay to stock pay. SPI increases the reliance of stock prices. Consistent with our earlier hypothesis, in addition to a decrease in the level of equity-based pay, we find high SPI firms linking compensation to stock prices in a more aggressive way. Another way SPI may affect equity-based compensation incentives is through performance-based pay. In performance-based pay, the relevant pay is conditional on whether certain performance targets are met. The performance metrics can be accounting measures, such as earnings and sales, but can also be the stock price-related. For example, the managers can only receive the relevant part of the pay if the stock price (return) is above a certain goal level. When SPI is higher, the stock price provides more useful information and should be relied more in the performance-based pay. We measure this reliance by the proportion of performance-based pay linked to stock price. The results are shown in Columns 5 and 6 in Table 9. The results show that both PIN and PSI have significantly positive effect on the proportion of performance-based pay linked to the stock price. It indicates that SPI increases the usage of stock price performance-based pay, which increases the pay-performance sensitivity. 8.2 Perquisites SPI may also improve the efficiency of compensation by cutting the perquisite component of the managerial compensation. Managers can enjoy the perquisites by misusing the resources of the firms, which destroys the shareholder value. SPI improves the observability 27

28 of managers actions, and thus makes managers less likely to demand higher perquisites. We expect SPI to reduce CEO perquisites. To test this, we regress the logarithm of perquisites on SPI measures. The results are shown in Table 10. Column 1 shows the result for PIN. The coefficient of PIN is negative and statistically significant at 5% level. The economic impact is also significant. One standard deviation increase of PIN decreases perquisites by 11%. Column 2 shows the result for PSI, which is consistent with the result for PIN in Column 1. One standard deviation increase of PSI decreases perquisites by 19%. The results indicate that SPI decreases the perquisites by increasing the observability of managers actions. Once again, SPI increases compensation efficiency while decreasing total compensation. 8.3 Pay for luck/skill In our last tests, we determine if SPI can increase compensation efficiency by improving the mix of pay for skill and luck. In an efficient compensation package, the manager should only be paid for their efforts or skills, rather than by luck. Higher SPI indicates that more firm-specific information, including the effort and skills of managers, are incorporated in the stock prices. Therefore, as SPI increases the observability of managers effort and skills, we expect SPI to decrease pay for luck and increase pay for skill. To test this hypothesis, we classify firms with PIN (PSI) in the top quartile by year as high SPI firms, and firms with PIN (PSI) in the bottom quartile by year as low SPI firms. We investigate the effect SPI has on pay for luck/skill in Table 11. We measure luck as the predicted firm stock return when regressing firm returns on industry returns, where the industry is defined by 2-digit SIC codes. We measure skill as the 28